Investing in AI

Govt should focus on enabling environment

The Union Cabinet last week cleared a Rs 10,371.92 crore package for the National AI (artificial intelligence) Mission, which will be focused on the creation of computing capacity in the country. This follows up on the earlier announcement that the government was considering setting up a three-tier computing structure. The details for the Mission budgeting, insofar as they are known, consist of viability-gap funding for the creation of high-end computing infrastructure by the private sector. A certain amount is also earmarked for investment in relevant startups as well as support to research and development (R&D) in academic institutions, including the establishment of platforms which will develop foundational models with capacities exceeding 100 billion parameters trained on datasets covering major Indian languages for sectors such as health care, agriculture, and governance. AI curation units (ACUs) will also be developed. The proposal also includes establishing an “AI marketplace”, designed to offer AI as a service and pre-trained models.

However, while all this would be useful and help the sector, the private sector is capable of finding the requisite investment for setting up computing infrastructure and developing the models and the marketplace, provided there is adequate policy support. Investment will be forthcoming for a high-growth area like AI. Indeed, the government can provide no more than seed funding for the ambitious targets the AI Mission outlines. The government should be focused more on providing an enabling environment. The AI ecosystem needs sensible, clear regulation and legislation, and enabling policy. This includes sensible Customs and tax policy for importing relevant hardware, depreciation for equipment, expensing guidelines for R&D in software, a push to encourage local manufacture of high-end equipment including semiconductors, and clear policy for investors in the AI ecosystem. Moreover, given that AI works on the basis of a vast amount of data, there is an urgent need to finetune the processing and storage of both personal and non-personal data as the sector grows. These are policy concerns and inputs only the government can address. Thus, it would be well advised to move forward in this direction as soon as possible.

As such, the Mission envisages viability-gap funding of around Rs 4,500 crore for setting up capacity aggregating around 10,000 high-end graphics processing units (GPUs). However, the allocation would only be sufficient to fund 1,000-1,500 in the way of high-end GPUs, though cost will reduce with scale. Another allocation of Rs 2,000 crore to funding startups can be reallocated to some other purpose, given the vibrancy and scale of the existing Indian startup ecosystem. An allocation of Rs 2,000 crore to the IndiaAI Innovation Research Centre for R&D into large foundational models, and the establishment of an IndiaAI Datasets Platform will certainly be useful, however. These could work as open cross-disciplinary platforms that support and service a multitude of public-sector and private entities. The commitment to an AI mission is commendable and so is the setting of ambitious targets. The world is moving fast in this context and India should not be left behind, particularly given its inherent advantages. However, policy formulation about the areas mentioned above is critical. It is important that the development and deployment of AI are under a sound and stable framework. Once there is clarity on the policy front, investment will flow into the sector.